You would absolutely not average them on the scanner because you cannot
correct for subject motion between the averages. You would first merge them
in the time dimension with fslmerge, run eddy_correct on the merged file,
split them back out with fslroi, and average them with fslmaths (add them
and divide by 2). There are example scripts posted on this list that
describe this process in detail.
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Adil Javed
Sent: Saturday, March 13, 2010 1:28 AM
To: [log in to unmask]
Subject: Re: [FSL] DTI imaging
Hi Matt,
This helps a lot.
Thanks so much.
To improve SNR, if I get 2 acquisitions of bo and 32 directions, do I
average these 2 acquisitions on the scanner or by using fslmeants or
fslinterleave?
aj
--- On Fri, 3/12/10, Matt Glasser <[log in to unmask]> wrote:
> From: Matt Glasser <[log in to unmask]>
> Subject: Re: [FSL] DTI imaging
> To: [log in to unmask]
> Date: Friday, March 12, 2010, 3:19 PM
> The way to go about this is the
> following:
>
> Register your FA to T1 image and generate a diffusion to
> structural
> transformation matrix. Invert this transformation
> matrix with convert_xfm
> to generate a structural to diffusion transformation
> matrix. Generate your
> lesion ROIs from your structural images (I am assuming you
> have registered
> your T2 to your T1), then move that ROI to diffusion space
> using nearest
> neighbor interpolation with FLIRT and extract your
> values. It is best to
> minimize the interpolation of your data in general, and
> this method allows
> you to avoid interpolating at all.
>
> Peace,
>
> Matt.
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]]
> On Behalf
> Of Adil Javed
> Sent: Friday, March 12, 2010 2:33 PM
> To: [log in to unmask]
> Subject: Re: [FSL] DTI imaging
>
> Hi Matt,
> just figured it out. Yes, I needed to brain extract
> my T1 before
> registration to FA maps, which is already brain
> extracted. Once I did that,
> then affine registration works fine.
>
> The registered FA-to-T1 looks slightly blurred. What I am
> trying to do is
> look at MS lesions on T2 scans (vox dimensions about 1x1x3)
> getting ROIs and
> then deriving FA/MD/tract values from the FA/MD/etc maps
> which were
> originally derived from 2x2x2 images.
> The question is would I lose any information on FA/MD/tract
> values by using
> interpolated registered FA-T2 maps for lesion analysis?
>
> Grateful for all your help,
> aj
>
> --- On Fri, 3/12/10, Matt Glasser <[log in to unmask]>
> wrote:
>
> > From: Matt Glasser <[log in to unmask]>
> > Subject: Re: [FSL] DTI imaging
> > To: [log in to unmask]
> > Date: Friday, March 12, 2010, 8:04 AM
> > That is very strange, it is common to
> > get good registrations with even
> > nonlinear registration with FA to T1. Are you brain
> > extracting your images?
> >
> > Peace,
> >
> > Matt.
> >
> > -----Original Message-----
> > From: FSL - FMRIB's Software Library [mailto:[log in to unmask]]
> > On Behalf
> > Of Adil Javed
> > Sent: Friday, March 12, 2010 1:25 AM
> > To: [log in to unmask]
> > Subject: Re: [FSL] DTI imaging
> >
> > Hi Matt,
> > Thanks for your comments, very educational and
> > helpful. When I transform FA
> > to T1 or T2, the only Model/DOF that works is
> Translation
> > (3 parameters
> > model). The others, including rigid body (6
> parameters) or
> > more (7-9) give
> > me very distorted images. Is registration by
> > translation only (3
> > parameters) good enough?
> > many thanks,
> > aj
> >
> > --- On Wed, 3/10/10, Matt Glasser <[log in to unmask]>
> > wrote:
> >
> > > From: Matt Glasser <[log in to unmask]>
> > > Subject: Re: [FSL] DTI imaging
> > > To: [log in to unmask]
> > > Date: Wednesday, March 10, 2010, 8:51 PM
> > > That is not the right way to go about
> > > things. You should leave the voxel
> > > dimensions at 2x2x2mm and then Register the FA to
> the
> > T1 to
> > > get the desired
> > > overlay. It is MUCH better to interpolate the
> > > calculated data than to
> > > interpolate the raw data and then calculate.
> This
> > is
> > > especially true if you
> > > want to do tractography. You can easily still
> > specify
> > > masks in structural
> > > space and then specify the structural to
> diffusion
> > > transform.
> > >
> > > Peace,
> > >
> > > Matt.
> > >
> > > -----Original Message-----
> > > From: FSL - FMRIB's Software Library [mailto:[log in to unmask]]
> > > On Behalf
> > > Of Adil Javed
> > > Sent: Wednesday, March 10, 2010 6:20 PM
> > > To: [log in to unmask]
> > > Subject: [FSL] DTI imaging
> > >
> > > Hi,
> > > for those who are experts in DTI:
> > > I am acquiring DTI images on 3.0T Phillips
> scanner
> > using 32
> > > directions, FOV
> > > 240, acquired matrix is 120 x 118 with acquired
> voxel
> > > dimensions of 2x2x2.
> > > In order to overlay DTI FA/MD/ or tracts on a
> FLAIR or
> > T1
> > > image, I need to
> > > reconstruct the matrix to 256 with voxel RECON to
> 0.94
> > x
> > > 0.94 x 2 so they
> > > align with my FLAIR or T1 images.
> > >
> > > Question: is this ok or do you see any
> problems
> > with
> > > this approach? By
> > > reconstructing voxels down to 0.94 x 0.94 x 2
> after
> > they
> > > are Acquired in
> > > 2x2x2 size, am I losing any information?
> > >
> > > many thanks for your help.
> > >
> > > adil
> > >
> >
>
|